7 Things That Can Go Wrong With Generative AI 7 Things That Can Go Wrong With Generative AI
Over the last couple of years, we’ve seen firsthand how generative AI is capable of creating content that is almost indistinguishable... 7 Things That Can Go Wrong With Generative AI

Over the last couple of years, we’ve seen firsthand how generative AI is capable of creating content that is almost indistinguishable from human-generated work. As the technology has scaled, it has shown that it has the power to revolutionize industries, from entertainment to education. But as Uncle Ben said, “With great power comes great responsibility.” That’s because the development and use of generative AI are fraught with complexities and challenges that must be navigated with care and consideration.

So let’s take a trip and delve into the key issues associated with generative AI, emphasizing the importance of responsible AI practices.

In-Person and Virtual Conference

April 23rd to 25th, 2024

Join us for a deep dive into the latest data science and AI trends, tools, and techniques, from LLMs to data analytics and from machine learning to responsible AI.


Biased Outputs

One of the most pressing concerns that has many worried with the rise of generative AI is the issue, and/or issues related to biased outputs. So a quick recap: AI models learn from vast datasets, and if these datasets contain biases—whether related to gender, race, or any other factor—the AI will inevitably learn and replicate these biases in its outputs. And here’s the kicker, this can lead to discriminatory content and perpetuate harmful stereotypes, undermining the inclusive potential of AI technologies. Addressing bias requires a concerted effort to curate diverse and representative datasets and develop algorithms that can identify and mitigate bias in AI-generated content.

This can often also require humans working with the AI, prompt engineers, NLP engineers, and other cross-disciplinary professionals to not only identify these biases but also formulate solutions.

Hallucinations and Misinformation

When people think about the potential harms of AI, this is often the first thing that comes to mind. That’s because generative AI models, for all their sophistication, can sometimes produce what is known as “hallucinations”—outputs that are entirely fabricated or grossly inaccurate. Since OpenAI’s ChatGPT went live, there are multiple examples of this hitting the internet. Sometimes funny in nature, but other times potentially dangerous and/or harmful. These instances of misinformation can have serious consequences, particularly when AI is used in sensitive contexts such as news generation, academic research, or legal analysis.

This is why when people use generative AI to generate new content, it is often advisable to do a quality check just in case. So why can’t AI just do it right the first time? Well, the thing is that ensuring the reliability of AI-generated content requires robust validation mechanisms and ongoing oversight by human experts, to prevent the spread of falsehoods and maintain the integrity of information. In short, it’s still a work in progress that likely won’t have an end anytime soon.



Now we’re entering the realm where technology can easily be used for harm. As many know, the ability of generative AI to produce deepfakes—hyper-realistic fake images, videos, or audio recordings—presents a double-edged sword. On one hand, this technology can be used for creative and beneficial purposes, such as in filmmaking or virtual reality. We’ve seen this recently in the new Indiana Jones movie which used deep fake technology to help “deage” star Harrison Ford for one last adventure. 

But on the other hand, it poses significant risks to privacy, security, and trust, as deepfakes can be used to create convincing misinformation or impersonate individuals for malicious purposes. There are already instances of black hat actors using deepfake technology to fool people that their family members need money or have been kidnapped. This is why it’s critical to combat the negative impacts of deepfakes necessitates advanced detection technologies, along with legal and regulatory measures to protect individuals and societies from their harmful effects.

Ethical Concerns

This isn’t a completely abstract concern. That’s because the development and deployment of generative AI raise profound ethical questions that can be easily overlooked due to the excitement generated by the work models produce. These range from the impact on employment and the economy to the moral implications of creating AI that mimics human behavior. It’s easy to understand why many are claiming that AI will have a greater impact on human society than the industrial revolution. This kind of shake-up can be a mixed bag of results that can hold both prosperity and harm.

One way to limit harm is by ensuring that AI serves the public good and involves engaging with diverse stakeholders, including ethicists, policymakers, and the wider community, to debate and define the ethical boundaries of AI innovation.

Stolen Data

What provides these models the ability to do what they’re doing? It is a vast amount of data that they use to learn or improve existing abilities. But often the need for this amount of data can pose significant privacy and security risks, particularly when personal or sensitive information is used without consent. Most of this is due to the fact that most training data is created by scraping the World Wide Web. This method makes it impossible to get the consent of every Internet user and has the potential to bring out other issues. 

This has created a demand from users to find ways of protecting against data theft while nations are looking to impose regulatory guardrails that companies would need to meet in order to ensure they’re in compliance with data protection regulations. As AI continues to grow, this will become even more crucial for maintaining trust in AI technologies and safeguarding individual rights. 

In-Person Data Engineering Conference

April 23rd to 24th, 2024 – Boston, MA

At our second annual Data Engineering Summit, Ai+ and ODSC are partnering to bring together the leading experts in data engineering and thousands of practitioners to explore different strategies for making data actionable.


Copyright Issues

Related to some issues we’ve just visited, as AI-generated content becomes increasingly indistinguishable from human-created work, copyright issues will come to the fore. Determining the ownership of AI-generated content and the rights to distribute, modify, or profit from it involves navigating a complex web of intellectual property laws. Not only that, but the training data containing the work of artists could bring companies into the court systems as artists look to exercise their copyright protections. Clear guidelines and legal frameworks are essential for fostering innovation while protecting the rights of creators and innovators.

Legal Issues

Though similar to some issues discussed above, the legal landscape surrounding generative AI is in flux, with many areas of law yet to catch up with the rapid pace of technological advancement. From liability for AI-generated content to the use of AI in decision-making processes – particularly within industries that have very robust regularity requirements, such as finance, banking, and others – legal professionals and policymakers must work together to develop regulations that balance innovation with accountability and protection for individuals and communities.

What’s next?

As you can see, even though generative AI can be seen as the new golden goose, there are still plenty of bumps in the road and issues that must be addressed in order to maximize the potential benefits of gen AI against any possible harm. Now if you want to keep up on the latest in generative AI, development, techniques, risks, frameworks, and more, then you’ll want to head over to ODSC East 2024

At East, you’ll be face to face with those pushing the boundaries of generative AI and sink your teeth into the best the community has to offer. So what are you waiting for? Get your pass today, and be ahead of the pack!

Here are some relevant sessions coming to the ODSC East 2024 Responsible AI Track:

  • How AI Impacts the Online Information Ecosystem
  • Resisting AI
  • Social and Ethical Implications of Generative AI
  • Advancing Ethical Natural Language Processing: Towards Culture-Sensitive Language Models
  • Guardrails for Data Teams: Embracing a Platform Approach for Workflow Management
  • HPCC Systems® for Social Good – Safe Havens!
  • How to Scale Trustworthy AI
  • Trust, Transparency & Secured Generative AI
  • AI and Society
  • Making AI recommendations Human-centric

And relevant sessions coming to the ODSC East 2024 Generative AI Track:

  • Intro to the ChatGPT API
  • Generative A.I. with Open-Source LLMs: From Training to Deployment with Hugging Face and PyTorch Lightning
  • Multimodal Retrieval Augmented Generation
  • Everything About Large Language Models: Pre-training, Fine-tuning, RLHF & State of the Art
  • Deploying Trustworthy Generative AI
  • Stable Diffusion: Advancing the Text-to-Image Paradigm
  • Aligning Open-source LLMs Using Reinforcement Learning from Feedback
  • State-of-the-art Open Source AI with Hugging Face
  • Graphs: The Next Frontier of GenAI Explainability
  • Beyond Theory: Effective Strategies for Bringing Generative AI into Production
  • Generative AI, AI Agents, and AGI – How New Advancements in AI Will Improve the Products We Build
  • Generative AI for Social Good
  • Generative Modeling in Quantitative Finance
  • Leveraging RAG and Multi-Agent LLM Systems for Automation of Knowledge Synthesis
  • Mastering PrivateGPT: Tailoring GenAI for your unique applications
  • Harnessing GPT Assistants for Superior Model Ensembles: A Beginner’s Guide to AI Stacked Classifiers
  • How to Defend Against Weaponized Generative AI
  • The Value of A Semantic Layer for GenAI
  • How to Rigorously Evaluate GenAI Applications
  • Implications of GenAI for Legal Services In the Americas and the APAC Region
  • Programming LLMs for Business Applications is Way Better Than ‘Tuning’ Them
  • Generative AI


ODSC gathers the attendees, presenters, and companies that are shaping the present and future of data science and AI. ODSC hosts one of the largest gatherings of professional data scientists with major conferences in USA, Europe, and Asia.